Visual analysis of character and plot information extracted from narrative text
The study of novels and the analysis of their plot, characters and other information entities are complex and time-consuming tasks in literary science. The digitization of literature and the proliferation of electronic books provide new opportunities to support these tasks with visual abstractions. Methods from the field of computational linguistics can be used to automatically extract entities and their relations from digitized novels. However, these methods have known limitations, especially when applied to narrative text that does often not follow a common schema but can have various forms. Visualizations can address the limitations by providing visual clues to show the uncertainty of the extracted information, so that literary scholars get a better idea of the accuracy of the methods. In addition, interaction can be used to let users control and adapt the extraction and visualization methods according to their needs. This paper presents ViTA, a web-based approach that combines automatic analysis methods with effective visualization techniques. Different views on the extracted entities are provided and relations between them across the plot are indicated. Two usage scenarios show successful applications of the approach and demonstrate its benefits and limitations. Furthermore, the paper discusses how uncertainty might be represented in the different views and how users can be enabled to adapt the automatic methods.